Lighting Design and Simulation Knowledgebase

Lighting Design Glossary

Diese Seite in Deutscher Sprache | Contact | Sitemap | Search

Image Filtering

(Term of computer graphics)

Digital images can be processed in a variety of ways. The most common one is called filtering and creates a new image as a result of processing the pixels of an existing image.

Each pixel in the output image is computed as a function of one or several pixels in the original image, usually located near the location of the ouput pixel. If the function used does some kind of interpolation (eg. linear, cubic or gaussian), then the result will look smoother than the original, but care needs to be taken that the ouput values are not computed from too many input pixels, or the resulting image may get blurred. The most common purpose for this interpolation is antialiasing.

In the Radiance software, high quality images are rendered by rpict(1) in higher resolution than desired, using a certain amount of pixel jitter. The resulting image, which will look a little rough in some areas, is then filtered to a lower resolution using gaussian interpolation with the pfilt(1) program. This process has the same effect as the oversampling used by other software, but offers much more control, since the filtering parameters can be optimised for the allready rendered image.

There are other filtering methods available, which are used to adjust the dynamic range of the image as with pcond(1), in order to adapt it to the display capability of the output medium, eg. hardcopy prints. In this process, the brighntess distribution in the whole image will influence the value of every pixel, which makes it possible to optimise the exposure and gamma curve of the image in a way that corresponds to a persons subjective impression when viewing a scene.


jittered sampling
Radiance software
English    German
image filtering    Die Bildfilterung
Copyright © 2004-2013